1. RESEARCH METHODS Lecture 32

2. The parts of the table 1. Give each table a number. 2. Give each table a title. 3. Label the row and column variables, and give name to each of the variable categories. 4. Include the totals of the columns and rows. These are called as marginals. 5. Each number or place that corresponds to the intersection of a category for each variable is cell of a table. 6. Missing information to be given under the table.

3. Percentaging Researchers convert raw count tables into percentages to see bivariate relationship. Three ways to percentage a bivariate table: by row, column, and for the total. Percentages by row and column are often used and show relationship

4. Right way to percentage Percentage by row or by column. Either could be appropriate. Decision based on researcher’s hypothesis. Age affects attitude towards women empowerment. Look at the table, see where is the independent variable. See in which direction its values are being added up. Percentage in the same direction

5. Reading a table Once we know how table is made, reading it and figuring out what it says are much easier. Look at the title of table, the variable labels and background information. Look at the direction in which percentages have been computed – in rows or columns

6. Read percentaged tables to make comparisons Comparison are made in the opposite direction from that in which percentages are computed. Compare across rows if the table is percentaged down the column.

7. How to figure out relationship? If no relationship: Percentages look approximately equal across row or columns. A linear relationship looks like larger percentage in diagonal cells.

8. Linear Positive relationship Linear Negative relationship X X Y Y

9. Linear relationship Table 4: Age by attitude towards women. empowerment. Age (in years). Level ofunder 4040 –60 61 +Total attitudeF.%F. % F % F % Hi Favorable600 60 300 30 200 20 1100 37 Med. Favorable 300 30 500 50 250 25 1050 28 Lo Favorable 100 10 200 20 500 50 850 28 Total 1000 100 1000 100 1000 100 3000 100 Larger percentages in the diagonal cells

10. If there is curvilinear relationship, the largest percentages form a pattern across cells e.g. the largest cells might be the upper right, the bottom middle, and the upper left.

11. Curvilinear

12. It is easier to see relationship in a moderate sized table (9 cells) where most cells have some cases (at least 5 cases)

13. A simple way to see strong relationship is to circle the largest percentage in applicable row or column and see if a line appears Table 4: Age by attitude towards women. empowerment. Age (in years). Level ofunder 4040 –60 61 +Total attitudeF.%F. % F % F % Hi Favorable600 60 300 30 200 20 1100 37 Med. Favorable 300 30 500 250 25 1050 35 Lo Favorable 100 10 200 20 500 850 28 Total 1000 100 1000 100 1000 100 3000 100 60 50

14. Is the relationship genuine? Eliminate alternative explanations – explanations that can make the relationship spurious. Experimental researchers do this by choosing a research design that physically controls potential alternative explanations for results. Control prior to the start of experiment

15. In non-experimental research: A researcher controls for alternative explanations with statistics. Measures possible alternative explanations with control variables Examines the control variables with multivariate tables and statistics and decides whether a bivariate relationship is spurious.

16. Control for third factor Two variable table. Relationship between age of people and attitude towards women empowerment. To see the spuriousness of X and Y introduce third variable i.e. gender. Control the effect of gender i.e. the effect of gender is statistically removed. After control, does the bivariate relationship still persist?

17. Control for gender Under each category of male and female, negative relationship between age and attitude persists. Relationship is not spurious. If bivariate relationship weakens, or disappears, then age is not the factor affecting the attitude. Difference in attitude is due to gender.

18. Statistical control A measure of association tested for its genuineness by controlling third variable. Researchers are cautious in their interpretations. Look for net effect. Go for trivariate percentaged table or multiple regression

19. Trivariate Table Test the alternative explanation. 3 rd factor. Control for third variable. Make a trivariate table. It has a bivariate table of XY for each category of control variable. The new tables may be called partials. No. of partial depends upon the No. of categories of control variable. Partial tables look like bivariate tables, but use subset of the cases. Break apart a bivariate table to form partials

20. Limitations Difficult to interpret if control variable has more than four categories. Total number of cases may be limiting factor because cases are divided into cells in partials. Thinning out of data. On average 5 cases per cell recommended.

21. Partial table for males Age (in years) Level of. Under 40 40—60 61+ Total. Attitude F % F % F. % F. %. High 300 60 200 33 30 6 530 33 Medium 140 28 270 45 120 24 530 33 Low 60 12 130 22 350 70 540 34 Total 500 100 600 100 500 100 1600 100

22. Partial table for females Age (in years) Level of.Under 40 40—60 61+ Total. Attitude F % F % F. % F. %. High 350 70 200 50 20 4 570 41 Medium 150 30 150 38 220 44 520 37 Low - - 50 12 260 52 310 22 Total 500 100 400 100 500 100 1400 100

23. Replication pattern When the partials replicate or reproduce the same relationship that existed in bivariate table prior to control Control has no effect.

24. The specification pattern When on partial replicate the same relationship but others do not. The researcher can specify in which partial there is strong relationship and where it is not.

25. The suppressor variable pattern When the bivariate table suggests independence of X and Y but the relationship appears in one or more partials. The control variable is suppressor – the true relationship appears in partials.

26. Multiple regression analysis